Periconceptional and First Trimester Ultraprocessed Food Intake and Maternal Cardiometabolic Outcomes

Characterizing diet by the nature and extent of processing has gained recent attention as a highly translatable framework for nutritional recommendations (1,2). Ultraprocessed foods (UPFs) are hyperpalatable industrial formulations containing chemicals, such as additives, and little to no intact or unprocessed foods (3). This category includes not only recognized dietary risk factors, such as sugar-sweetened beverages, but also understudied subgroups such as ready-to-eat meals. UPFs are ubiquitous, representing nearly 60% of calories consumed in the U.S. (4).

In a human feeding trial comparing an ultraprocessed diet to an unprocessed diet matched for presented calories and other dietary constituents, the ultraprocessed diet led to higher ad libitum energy intake and weight gain (5). This finding suggests that the influence of UPFs goes beyond their nutrient density; they are hypothesized to disrupt the gut-brain axis satiety signaling, promoting unregulated intake and subsequent weight gain (5). Despite this established link outside of pregnancy, few studies have examined whether higher UPF intake is associated with an increased risk for excessive weight gain in pregnancy (6–8). These studies were limited by small nonrepresentative samples as well as issues related to temporality.

Prospective studies in the general population have also linked high UPF intake with type 2 diabetes (9) and hypertension (10), through pathways other than excess energy intake and weight gain. Additives commonly found in UPFs have been shown in animal studies to inhibit insulin signaling (11), and UPF consumption has been associated with increased urinary concentrations of phthalate metabolites, which are known endocrine disruptors (12). Pregnancy is a period of cardiometabolic susceptibility due to naturally heightened insulin resistance. However, few studies have examined the association of UPFs with maternal outcomes, including gestational diabetes mellitus (GDM) (8,13). Further, although GDM and hypertensive disorders of pregnancy (GHTN) share aspects of their etiology, such as insulin resistance (14), to our knowledge, no prior studies have examined the association of UPF intake with GHTN.

In a large, diverse, multisite, prospective cohort of U.S. pregnant women, we examined the association of UPF intake per the NOVA diet classification system (3,15) in periconception and the 1st trimester with weight gain, glycemic, and blood pressure outcomes. Our primary outcomes were gestational weight gain (GWG) by Institute of Medicine (IOM) criteria, GDM, and GHTN. As secondary outcomes, we also examined associations with GWG rate, glucose challenge test 1-h glucose, and systolic and diastolic blood pressure trajectories across pregnancy.

The Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) Fetal Growth Studies–Singletons (2009–2013) was a prospective cohort study of U.S. pregnant women designed with the primary aim of establishing race/ethnicity-specific fetal growth standards (16). The cohort included 2,802 self-identified non-Hispanic White, non-Hispanic Black, Hispanic, and Asian/Pacific Islander women from 12 clinical sites. Chronic diseases and smoking were exclusion criteria for women without obesity (n = 2,334), with less restrictive exclusion for chronic conditions for women with obesity (n = 468). The study was approved by the institutional review boards of all clinical sites, data coordinating centers, and the NICHD. All women provided written informed consent prior to enrollment.

The analytic sample for the present report included 1,948 women who had a live birth and were enrolled in the study after the nutrition component was deployed. Analyses involving blood pressure trajectories were done in a subsample of 1,792 women with the data available from prenatal care records.

Dietary intake of the 3 months prior to enrollment was assessed using a validated 144-item food frequency questionnaire (FFQ) at 8–13 weeks of pregnancy (17). Food and beverage items in the FFQ were categorized as UPF following the NOVA classification system, which categorizes foods into four groups based on the nature and degree of processing (3). The UPF category includes items such as soft drinks, cookies, candy/confectionery, sweet or savory packaged snacks, and ready-to-eat/heat meals. Previous studies have derived UPF variables from FFQs designed to assess usual diets in the U.S. (7,18,19), Spain (10,13,20), and Brazil (21,22).

Due to the limited nature of food preparation data from diet assessed by an FFQ, some items were not easily classifiable by NOVA criteria. We therefore defined a strict UPF exposure variable including only items that were clearly in the UPF category. As a sensitivity analysis, we also defined a broad UPF exposure variable that additionally included items that were likely in the UPF category. For example, juices are considered unprocessed/minimally processed if they are “fresh or pasteurized fruit or vegetable juices (with no added sugar, sweeteners or flavors)” (23), while packaged juices are considered UPF. We therefore included items such as “orange/grapefruit juice” in the FFQ under the broad but not the strict UPF exposure variable given that information on specific processing was not available. We identified 34 FFQ items as UPF using the strict criterion and 54 items using the broad criterion (Supplementary Table 1).

Daily consumption of UPF items on the FFQ was calculated by multiplying the frequency of intake with the serving size that was directly indicated on the FFQ or estimated from the U.S. Department of Agriculture Food Data central database (fdc.nal.usda.gov). Items were summed to calculate the quantity (g/day) of UPF intake by the strict and broad definitions. Implausible dietary data, defined as total energy intake <600 kcal/day or >6,000 kcal/day were set to missing and later addressed using multiple imputation. The Healthy Eating Index-2010 (HEI-2010) was used as an indicator of diet quality, reflecting adherence to the Dietary Guidelines of Americans during the study period (24). Major food and food group servings were defined according to the MyPyramid Equivalents Database (25).

Women self-reported their prepregnancy weight at enrollment. Women’s weight as pregnancy progressed was measured at study visits using a standardized protocol (26) and was abstracted from prenatal care clinic medical records. Second and third trimester weight gain was calculated, along with the rate of weight gain in both trimesters (kg/week), and categorized per IOM recommendations (inadequate, adequate, and excessive) (27). Recommended ranges of 2nd and 3rd trimester weight gain for women with underweight, normal weight, overweight, and obesity are 0.44–0.58 kg/week, 0.35–0.50 kg/week, 0.23–0.33 kg/week, and 0.17–0.27 kg/week, respectively. Weight gain rates below and above the recommended ranges were considered inadequate and excessive GWG, respectively.

Women underwent the standard of clinical care, which included a nonfasting 1-h glucose challenge test, followed, if positive, by an oral glucose tolerance test (OGTT). For women with available OGTT results (n = 376), GDM was defined based on Carpenter and Coustan criteria, as endorsed by the American College of Obstetricians and Gynecologists and the American Diabetes Association, of having at least two high plasma glucose measurements (fasting: ≥5.3 mmol/L, 1-h: ≥10.0 mmol/L, 2-h: ≥8.6 mmol/L, 3-h: ≥7.8 mmol/L) (28). Women were also assumed to have GDM if medication treatment for GDM was recorded in their discharge diagnosis. One-hour glucose (mg/dL) after a universal 50 g oral glucose challenge screening test was also used as a continuous measure of glycemia.

Presence and severity of GHTN was determined by a physician based on the standard of care at each study clinic. Women were categorized as having 1) no hypertension; 2) mild gestational hypertension or unspecified hypertension; or 3) severe gestational hypertension or severe/mild preeclampsia (29). Continuous systolic and diastolic blood pressure measurements were also abstracted from clinic medical records with up to 28 repeated measures across pregnancy.

At enrollment, research nurses interviewed women to obtain sociodemographic characteristics (age, race/ethnicity, income, education, student/employment status, marital status, insurance coverage) and reproductive history (parity). Women also self-completed a validated physical activity questionnaire regarding their activities in the past 12 months from which moderate and vigorous physical activity (MET/min) was calculated (30). As women enrolled in the study at 8–13 weeks’ gestation, the first measured weights reflect a range of early pregnancy weight gain or loss. Therefore, prepregnancy BMI was calculated from self-reported weight and measured height. Self-reported prepregnancy weight has been shown to have high correlation with measured weight (31). Average sleep duration (h/day) during pregnancy was self-reported in a questionnaire.

UPF intake (g/day) was ranked into quartiles, and characteristics of women across quartiles of UPF intake were described. Per Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) recommendations (32), significance tests were not included in descriptive tables.

Prospective associations of UPF intake with continuous, binary, and multinomial outcomes were evaluated using linear, log-binomial, and logistic regression, respectively, using the lowest quartile of UPF intake as the reference group. Tests of linear trend were conducted by using the median value for each quartile and fitting it as a continuous variable in the applicable regression models.

Systolic and diastolic blood pressure trajectories from repeated measures across pregnancy were estimated using linear mixed models with cubic splines to allow for potential nonlinearity with knots at the 25th, 50th, and 75th percentiles of gestational weeks (17, 28, 35 weeks). P values for the adjusted associations of UPF intake with blood pressure trajectories were estimated using a log-likelihood ratio test comparing models with and without the exposure. Trajectories for each quartile of UPF intake were plotted at the average distribution of covariates in the sample.

Covariates were selected a priori with the aim of having the most parsimonious models while accounting for established confounding variables. These included age, race/ethnicity, education, marital status, nulliparity, prepregnancy BMI, moderate to vigorous physical activity, sleep duration, and total energy intake for glycemic and blood pressure outcomes. Family history of diabetes was also a covariate for glycemic outcomes. Although UPFs are hypothesized to be dietary risk factors for cardiometabolic outcomes independently of their caloric contribution, the mechanism of association with weight gain would operate through energy balance. Therefore, models for GWG outcomes were not adjusted for total energy intake.

All primary analyses performed using the strict UPF exposure variable were repeated using the broad UPF exposure variable to assess whether potential misclassification in categorizing FFQ items into UPFs could result in differential associations with outcomes. Missing data, including implausible dietary data, were addressed using 20 iterations of multiple imputation assuming they were missing at random. All analyses were performed using SAS 9.4 software (SAS Institute, Cary, NC).

To evaluate associations in women at different levels of risk for GDM, we tested for interactions with family history of diabetes as well as with maternal age dichotomized as <30 years and ≥30 years.

Median UPF intake in the analytic sample was 1,540 g, with the range of intake in quartiles 1 and 4 being 44–1,128 g and 1,945–3,398 g, respectively. Women with higher UPF intakes were younger, of higher BMI, had higher physical activity, and were less likely to identify as Asian/Pacific Islander (Table 1). Women with higher UPF intake also had greater total energy intake, with a higher percentage of energy from fats (in particular, saturated fat) and a lower percentage of energy from protein, and were less likely to self-identify as vegetarian. Higher UPF intakes were also accompanied with lower diet quality per the HEI-2010 (quartile 4 vs. 1: 60.4 vs. 68.9). Distributions of most characteristics across quartiles of UPF intake using the broad criterion were similar (Supplementary Table 2).

Adjusted associations of UPF intake with gestational weight gain are presented in Table 2. A total of 492 (25.2%) and 699 women (35.9%) had excessive GWG in the 2nd and 3rd trimesters, respectively. Higher UPF intake using the strict criterion was associated with lower odds of excessive 2nd trimester weight gain in certain quartiles (quartile 2 vs. quartile 1: adjusted odds ratio [OR_adj] 0.65 [95% CI 0.44, 0.98]; quartile 3 vs. quartile 1: OR_adj 0.60 [0.39, 0.92]) without a significant linear trend (P-trend = 0.10). Higher UPF intake using the broad criterion was associated with lower odds of 2nd trimester inadequate weight gain (quartile 4 vs. quartile 1: OR_adj 0.63 [95% CI 0.44, 0.91], P-trend = 0.01) and 3rd trimester inadequate GWG (quartile 4 vs. quartile 1: OR_adj 0.60 [0.41, 0.89], P-trend = 0.05). Certain quartiles of UPF intake were also associated with lower odds of 2nd trimester excessive weight gain using the strict (quartile 3 vs. quartile 1: OR_adj 0.65 [95% CI 0.39, 0.92]) and broad (quartile 3 vs. quartile 1: OR_adj 0.60 [0.39, 0.91]) criteria, without a significant P-trend.

Differences in diastolic and systolic blood pressure trajectories by UPF intake were not clinically significant, with women in the highest quartiles of UPF intake by the strict criterion having slightly higher blood pressure across pregnancy (Fig. 1). For example, at gestational week 37, women in quartiles 4 and 1 had mean diastolic blood pressures of 69.9 mmHg and 68.6 mmHg, respectively (overall trajectory P = 0.034). Similarly, their mean systolic blood pressures were 114.5 mmHg and 112.4 mmHg, respectively (overall trajectory P = 0.007). UPF intake using the broad criterion was associated with diastolic and systolic blood pressure trajectories (P = 0.030 and P = 0.010, respectively) but was also not clinically significant.

Adjusted associations of UPF intake with glycemic outcomes and GHTN are presented in Table 3. A total of 85 women (4.4%) had GDM and 63 (3.2%) had severe hypertension or preeclampsia. UPF intake using the strict criterion was not associated with 1-h glucose (quartile 4 vs. quartile 1: β_adj = −0.95 mg/dL [95% CI −5.07, 3.17], P-trend = 0.60) or GDM diagnosis (quartile 4 vs. quartile 1: adjusted risk ratio [RR_adj] 0.99 [95% CI 0.46, 2.11], P-trend = 0.85). Women with higher intake of UPFs did not have greater odds of GHTN outcomes. There were also no significant associations of the broad UPF variable with glycemic outcomes and GHTN.

There was no difference in the association of UPF intake with risk of GDM according to women’s family history of diabetes (P-interaction = 0.30 for strict UPF, P-interaction = 0.10 for broad UPF). However, UPF intake using the broad criterion had a significant interaction with dichotomized age (P-interaction = 0.02). Among women <30 years of age, higher UPF intake trended toward a positive but nonsignificant and imprecise association with GDM (quartile 2 vs. quartile 1: RR_adj 0.94 [95% CI 0.27, 3.31]; quartile 3 vs. quartile 1: RR_adj 0.93 [0.25, 3.43]; quartile 4 vs 1: RR_adj 2.98 [0.89, 9.97]; P-trend = 0.05). Associations were not significant among women ≥30 years (quartile 2 vs. quartile 1: RR_adj 0.40 [95% CI 0.16, 1.00]; quartile 3 vs. quartile 1: RR_adj 0.68 [0.28, 1.67]; quartile 4 vs 1: RR_adj 0.35 [0.10, 1.25]; P-trend = 0.12).

In a prospective cohort of low-risk U.S. women with singleton pregnancies, we found that UPF intake in the periconceptional and 1st trimester period was not associated with higher excessive weight gain or adverse glycemic and blood pressure outcomes during pregnancy. Contrary to our hypothesis, higher UPF intake was not associated with higher odds of excessive GWG by IOM criteria; in fact, inverse associations with inadequate GWG were observed when using the broad UPF criterion. UPFs also did not appear to be a dietary risk factor for GDM. Moreover, differences in diastolic and systolic blood pressure trajectories across quartiles of UPF intake were not clinically meaningful and did not translate into higher odds of GHTN. These findings stand in contrast to the evidence in the general population and must be interpreted with caution due to the limited number of women with GDM (n = 85) and GHTN (n = 63 with severe hypertension or preeclampsia) in our cohort. If replicated in studies with greater statistical power, however, they may suggest that nutritional recommendations focusing solely on avoiding UPFs may not be effective for the prevention of these outcomes in the context of pregnancy.

Associations of UPF intake with weight gain in our cohort are more nuanced than in previous studies that reported strong positive associations with weight gain or obesity during pregnancy. In a clinical sample of pregnant women (n = 45) in Midwestern U.S., Rohatgi et al. (7) found a 1.33-kg increase in GWG for every 1% increase in energy from UPFs at 30–34 weeks. In a cross-sectional study of Brazilian pregnant women (n = 785) at 24–39 weeks, Sartorelli et al. (8) reported that higher UPF intake was associated with higher odds of maternal obesity per gestational age-specific BMI categories. Lastly, de Barros Gomes et al. (6) found that consumption of UPFs in the 3rd (but not 2nd) trimester was positively associated with weekly GWG in the 3rd trimester among Brazilian pregnant women (n = 259).

Our study builds on this previous set of work by including a large, multisite cohort with a prospective design improving generalizability and minimizing chances of reverse causality. We found that UPF intake by broad, but not strict, criteria was inversely associated with inadequate GWG in the 2nd and 3rd trimesters. This favorable association does not necessarily support the indiscriminate promotion of UPFs for meeting weight gain goals during pregnancy as these findings must be balanced with the potential risks for offspring health (7) and adverse outcomes outside of pregnancy (33). We also observed some associations of lower odds for excessive GWG, although there was not a consistent dose response across increasing quartiles. The biological plausibility of this association is questionable, and residual confounding cannot be ruled out. Given these sometimes contradictory associations across both ends of IOM categories and the inconsistency by strict versus broad criteria, we are only able to conclude that the expected unfavorable association of UPFs with higher odds of excessive GWG was not observed in our cohort of low-risk pregnant women.

Our results also indicate that UPFs may not be independent dietary risk factors for GDM and GHTN. The dietary profile (e.g., high glycemic index and glycemic load) and chemical content (both as ingredients and in packaging) of UPFs are hypothesized to disrupt glucose metabolism (9,34). However, we did not observe an association between higher intake of UPFs and 1-h glucose or GDM. Likewise, Sartorelli et al. (8) found no association of UPFs with GDM. UPFs were also not associated with cardiometabolic biomarkers, including fasting glucose and insulin among pregnant women in the Rohatgi et al. (7) study. In a prospective cohort of pregnant women in Spain, an association of pregestational UPF intake with higher GDM risk was observed only among women aged ≥30 years (13). Although we detected an interaction of dichotomized age and UPF intake when using the broad criterion, associations with GDM were nonsignificant and imprecise. Though the high sodium-to-potassium ratio (34) of UPFs is also thought to adversely influence blood pressure, differences in blood pressure trajectories in our cohort were not clinically meaningful, and there were no associations with GHTN.

A limitation in estimating UPF intake from FFQs is the subjectivity involved in determining levels of processing for some items. For example, a study from Spain classified “breads (white and whole)” as processed (10), while a U.S. study classified breads as ultraprocessed, citing that most breads are unlikely to be homemade (7). Presently, this challenge seems unavoidable given that a notable degree of subjectivity also applies to classifying 24-h recall data (35), and FFQs (36) as well as other assessment tools (37) designed specifically to capture NOVA categories are just starting to be developed.

A major strength of our analyses is that we evaluated a broad UPF variable in addition to the strict UPF exposure to allow maximum comparison with previous studies and to evaluate the effect of potential misclassification. Reassuringly, our major conclusions were consistent across the two criteria for glycemic and blood pressure outcomes. Another strength of our study is that it is the first to examine UPF intake in relation to longitudinal blood pressure and diagnoses of GHTN. This is important given the mounting evidence for the association of preeclampsia with long-term maternal cardiometabolic risk (38). Our use of the 1-h glucose challenge test as a continuous measure of glycemia is a limitation. However, the more reliable measure, fasting blood glucose from OGTT, was available only in a small subsample of the cohort. Although the current study is the largest of its kind among U.S. pregnant women, the relatively low prevalence of GDM and GHTN in this low-risk cohort may have resulted in a lack of statistical power to detect associations of low magnitudes. Additionally, the findings are generalizable only to women without major chronic diseases at the start of their pregnancy. Lastly, although we adjusted for a number of important covariates, residual confounding may still be present.

In conclusion, in a prospective cohort of low-risk pregnant women, we found that UPF intake in periconception and the 1st trimester was not associated with higher excessive weight gain and adverse glycemic and blood pressure outcomes. Eliminating or reducing UPFs is being increasingly recommended for promoting cardiometabolic health in the general population (2,39). Our results must be interpreted in the context of the evidence on UPFs outside of pregnancy as well as the shortcoming of limited statistical power. If our findings hold true in cohorts with larger cases of GDM and GHTN, an implication may be that prenatal dietary advice should incorporate more specific targets than avoiding this one category of a food processing framework.

Acknowledgments. The clinical centers involved in data collection for the Eunice Kennedy Shriver National Institute of Child Health and Human Development Fetal Growth Studies were (in alphabetical order): Christina Care Health Systems, Columbia University, Fountain Valley Hospital, Long Beach Memorial Medical Center, New York Hospital, Queens, Northwestern University, University of Alabama at Birmingham, University of California, Irvine, Medical University of South Carolina, Saint Peters University Hospital, Tufts University, and Women and Infants Hospital of Rhode Island. C-TASC and The EMMES Corporation were the data coordinating centers that provided data and imaging support for this multisite study.

Funding. This work was supported by the Intramural Research Program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health (Bethesda, MD) contracts HHSN275200800013C, HHSN275200800002I, HHSN27500006, HHSN275200800003IC, HHSN275200800014C, HHSN275200800012C, HHSN275200800028C, and HHSN275201000009C.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.F.Y. drafted the manuscript. S.F.Y., S.N.H., S.L.M., J.L.G., K.L.G., C.Z., and J.G. contributed to the interpretation of the results and revision of the manuscript for important intellectual content and approved the final version of the manuscript. S.F.Y., S.N.H., S.L.M., and J.G. conceptualized the study design. S.F.Y., and J.L.G. analyzed data. S.N.H., K.L.G., C.Z., and J.G. contributed to data acquisition. S.F.Y., and J.G. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented as an abstract at the American Diabetes Association’s 81^st Scientific Sessions virtual meeting, 25–29 June 2021 and Best of ADA Virtual Program in June 2021.